Create a Custom AI Assistant + API in 10 Mins
HTML-код
- Опубликовано: 6 июн 2024
- Learn how to build your own AI assistant using OpenAI's Assistants API and how to access it via Python.
👉 Links
🔗 Code: github.com/pixegami/openai-as...
🔗 Documentation: platform.openai.com/docs/assi...
🔗 OpenAI Assistants: platform.openai.com/assistants
🔗 Data (PDF): www.eiu.com/n/campaigns/globa...
📚 Chapters
00:00 OpenAI Assistants API
00:58 Create an OpenAI Assistant
02:07 Adding Custom Data
04:03 Function Calling
05:58 Code Interpreter
06:57 Using an API Endpoint
10:03 Wrapping Up
#pixegami #ai
![TRASH or PASS! Eminem ( Houdini ) [REACTION!!!]](http://i.ytimg.com/vi/cB1WrZcMwMc/mqdefault.jpg)
Excellent Bro!!! Clear and concise👍
Thank you :) Hope you enjoyed it.
To the point, good quality. Subscribed.
Thank you!
Love your videos. Thanks for sharing 👍
Thank you, glad you enjoyed it!
Thanks for the video man. This helped me understand how I can transfer my custom-built OpenAI chatbot to my website/SaaS. Basically it's an interrelationship with VSCode and OpenAI and then once I'm happy I just use the API and chabot ID and somehow plug it into my website on the backend. At least that's what I think. We'll see.
Glad to hear it helped you get going. There's a lot of different ways to build what you want, so the best thing is just to go ahead and try it out like you are doing. Good luck!
Another awesome video! Thanks!
Glad you enjoyed it! Thank you :)
so concise and clear! thank you!
Glad to hear it was useful!
Good tutorial! Thanks! 👍
Thanks!
I see how much I have been over complicating this. Thanks.!
Glad it helped!
great work man! thanks for sharing
You're welcome! Glad you liked it!
Wow you are amazing. Everything is well explained and completed. I am so impressed.
Thank you so much! I really appreciate the feedback :)
Great video. I will try it a little bit.
Thank you. Good luck!
Dude you are much better looking than Sam. Use your own pic and be proud :-)
I'm flattered. Thank you!
Such a simple and easy eplaination. I wish you were my college professor.
Thank you :)
Very good video. Thank you 👌
Glad you liked it!
thanks a lot it's really helpful, can you make part 2 for build a custom UI using streamlit with AI Assistant
Awesome thank you! @pixegami Do you know how to print screen what tool the assistant is using while answering your query? Assuming there are multiple tools available to the assistant. Thanks.
Hmm, I think maybe if you enable the verbose setting in the assistant, you might be able to get more details out of it? Not quite sure, but I think it'll probably be somewhere in the OpenAI docs. I'd be shocked if they didn't expose that feature :P
great video
Thank you :)
Great video! What steps would need to be included to allow the Assistant to use data from a 3rd party API? I’m not a coder and hope there is a way to connect to an external API from the Playground using Functions tool?
That's probably a bit more complex, I think currently you do need a bit of coding to get that working.
So the OpenAI assistant will just give you the raw data to send to the 3rd party API. You'll have to write (or generate) some code to use that input and make the call yourself, then you can decide what to do with the response (e.g. feed it back to the assistant, or do something else with it).
Perfect
Nice video!!!!!
Thank you!
Any suggestions on fast cheats for creating these JSON or Yaml schemas? Something in VS code, feeding question into GPT? Havent been able to get them to automatedly generate these template styles
Hmm, for something like this, I think since it's usually a "one-off" cost to developing an assistant, it's probably best to do the schemas by hand so you can validate them and make sure they are to your spec. You'll always get variable results with LLMs to help on schemas like this that are not widely used or documented in its training data.
Hey - what do you use to edit the gradient circle around your camera?
I just prepare the graphic in photoshop (it's just a gradient circle), then use OBS to record. I put a circular mask around my camera image, and just stick that on top of the circle.
great video,i ve been struggling with the API connexion with my assistant,do you have an equivalent code for R?
Ah, sorry I don't have a ton of experience with R. :(
Can you force it only to look into your data for its responses? Very nice video. Thank you.
I'm not sure how possible that is from the Assistant tooling itself. But (either using this or using the something like Langchain and RAG directly) you could also implement your own validation that the data it uses must closely relate to the material you've given it.
Great video. I'm still new. How is this different from asking chat GPT to summarize a document for me?
It lets you configure the GPT with a bit more context upfront, so if you wanted to build an app or a UI on top of this for other people to use, this gives you a head start.
so this is basically doing a task which langchain or llamaindex is doing?
Sort of. The UI/UX is a little easier and more managed. But you can definitely also implement something like this yourself with Langchain.
sorry im new to OAI assistant API..i have already had 'credentials' from my custom GPT in GPT store. how can i connect it to my separate app with assistant API? not recreating the instruction on assistant API but just connect it to custom GPT in GPT store that i have already created earlier
I think the Custom GPT and the Assistant APIs are slightly different products in OpenAI's backend, so you'll probably have to re-implement your Custom GPT as an Assistant first, then use that as an API in your app.
Does the ai assistant have the ability to retain memory?
There is no Retrieval in the tools on my screen. Can you guess what the problem is?
But the GPT's in the gpt store are capable of accessing external API's, why aren't the assistant capable of doing so ? Do you know any alternate models which allow for external API access ?
I'm not sure why they allowed the GPT store to do that (do they really!?), but not custom assistants. But I guess that's something OpenAI expects you to handle in your own server side logic.
ImportError: cannot import name 'OpenAI' from 'openai'
This is a library compatibility problem. Any fix? ChatGPT has me running in circles literally with the same 2 set of fixes. Thanks. Still a good video!
The problem was cause by VS Code IDE. When I created a new venv, ran the main.py as a script, it was fine. So I need to either new a new IDE, or find the config issue in my VSC. Thanks.
It's probably that you need to select the Python environment to use. By default, I think VSCode might not know which Python environment you need to use. Don't worry, it's quite a common snag that I get caught on a lot as well.
Here's how to fix it: code.visualstudio.com/docs/python/environments
Is there any way to personalize assistant on mobile app?
I guess you can write a mobile app as a frontend to interact with your own AI assistant API?
im unable to save documents for reference.. any idea?
Hmm, did you check if the file format of the doc (e.g. PDF, text, etc) and the file size is supported by OpenAI?
May I ask wht I only have GOT3.5 turbo, no Gpt 4. I also paid.
Hmm, do you have the premium membership subscription? That's all I have and it shows up for me. I don't know if there's any geographical restrictions... (I'm in Australia).
Hey @willyu7515, all you need to do is add some credits to your account (minimum $5). It's terrible that they don't mention it, but I had the same problem, and as soon as I added some money, the GPT 4 popped up. Hope it helps!
CAn I ajust Tokens and temperature?
Hmm, I don't think the Assistants API allows you to do that. If you want a finer level of control, it might be best to implement your own agent using the model directly and something like Langchain library for utility.
but assistant api still not have history right ?
I think with "threads" you can have history for that thread. You'll just have to store the thread ID to retrieve it?
Would you like to be a Judge?
1. CodeCraft Duel: Super Agent Showdown
2. Pixel Pioneers: Super Agent AI Clash
3. Digital Duel: LLM Super Agents Battle
4. Byte Battle Royale: Dueling LLM Agents
5. AI Code Clash: Super Agent Showdown
6. CodeCraft Combat: Super Agent Edition
7. Digital Duel: Super Agent AI Battle
8. Pixel Pioneers: LLM Super Agent Showdown
9. Byte Battle Royale: Super Agent AI Combat
10. AI Code Clash: Dueling Super Agents Edition
As far as I understand, it is impossible to call external API directly from the chat GPT agent. It is only possible to call an assistant from the Python code locally on a PC.
Yup, that's my understanding as well.
doesn't it works with excel files?
I'm not sure, I think it does? But if not, you can probably export the Excel file into a CSV (which I'm pretty sure it should work with).
Can you do a django tutorial ?
Thanks :) Django is requested quite a lot so I've scheduled in to work on a tutorial in the next 3 months or so :)
2:10
Unfortunately it's not free, it reponds with quota exceeded.
Yup, I think it might require an OpenAI account. Have you tried changing it to a different (older) model?
The way they constructed this API, with runs and threads and messages - it's hilariously unintuitive and bad. Almost as bad as a Google API - I mean not that bad but still.
It's like they purposely decided to make it incomprehensible.
I agree it was incredibly unintuitive to use. But designing APIs like this will always have a lot of trade-offs, and I think the designers here were really positioning it for high-scaling, async-first use cases.
*most livable *most boring :D
Well... 🤷♂️
Thank you for this great tutorial🤍
why it doesn't show me chatGPT4 and only chatGPT3.5? and can i upload images to the assistant?
you need to get the premium subscription for chatgpt-4 and i think it does accept images
Yup, it seems as someone else mentioned, you might need to have the OpenAI premium subscription to access GPT-4.